Compliance guide · Algorithmic trading

SEBI's retail algo-trading framework, in plain English

~15 minute readLast updated 2026-07-05Educational, not legal advice

The short answer

SEBI's circular of 4 February 2025, titled Safer participation of retail investors in Algorithmic trading, brings retail API and algo trading under a registration and accountability regime rather than banning it. Orders placed through a broker's API above a specified order-frequency threshold, widely reported as around ten orders per second, are treated as algorithmic and must be registered and tagged with a unique algo ID at the exchange. The broker becomes the principal responsible for the algo and API providers it empanels, API access is hardened, disclosed white-box logic is separated from opaque black-box strategies, and the framework is phased to a go-live for all brokers on 1 April 2026.

For years, running a Python strategy through a broker's API sat in a regulatory grey zone: technically possible, lightly supervised, and surrounded by a market of third-party algo and signal sellers that no one was really answerable for. This circular ends the grey zone. It does not outlaw retail algos; it makes them identifiable and puts a regulated intermediary on the hook for them.

1. What the circular actually does

Start with the problem it addresses. Broker APIs and no-code order platforms made algorithmic execution accessible to ordinary retail traders through the early 2020s. Around that access grew a cottage industry of third-party algo providers and signal sellers, many of them connecting to markets with little formal oversight, some of them selling opaque strategies on performance claims a buyer had no way to verify. From the regulator's seat, three things were unsatisfactory at once: automated orders were hard to trace back to a responsible party, the providers pushing them sat outside the regulated perimeter, and the retail buyer bore the risk of a black box they could not inspect.

The 4 February 2025 circular, reference SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/0000013, answers all three by construction rather than by prohibition. It defines when API-routed activity counts as algorithmic, requires that such orders be registered and tagged so each one is traceable, and reassigns responsibility for the whole chain to the broker as a regulated principal. Its own stated purpose is safer participation, with stock brokers and stock exchanges taking on defined roles in risk management. The word to hold onto is accountability: the framework is less about what you may trade and more about who answers for the automated order once it leaves your machine.

2. The classification test: when an order becomes an algo

The framework needs a bright line between a person clicking buy through an interface and a program firing orders at machine speed, because the two carry very different systemic risk. It draws that line at an order-frequency threshold. Below the threshold you are treated as an ordinary API user; above it, your order flow is treated as algorithmic and pulled into the registration-and-tagging regime.

The figure most widely reported for that threshold is around ten orders per second, measured on a per-client basis within a short window. The precise value, and exactly how it is counted, are operationalised through the exchanges and the industry standards process rather than frozen in the circular text, so the sensible posture is to treat ten orders per second as the working line while confirming the current number with your own broker and exchange. What matters conceptually is that the trigger is a rate of automated order submission, not the source of the trading idea: a fast execution program crosses it, a human placing occasional discretionary orders does not.

How an order is classified as ordinary or algorithmic A retail order enters through a broker's API. A decision checks whether the order rate exceeds about ten orders per second. Below the threshold the order is ordinary. Above it the order is algorithmic: the broker registers and tags it with a unique algo ID and the exchange records it for oversight. The classification test Retail order via a broker's API Order rate above ~10 per second? NO YES Ordinary API order no algo registration Algorithmic order enters the regime Broker registers and tags a unique algo ID exchange records it Traceable to strategy and provider The threshold value is operationalised through the exchanges; treat about ten orders per second as the working line and verify the current figure.
The rate decides the regime. The test is mechanical and speed-based, not intent-based. Once an order is on the algorithmic side of the line, identification is built into the order itself through the unique algo ID, which is what makes every automated order traceable back to the strategy and provider that produced it.

3. The key provisions, at a glance

Beneath the headline, the framework is a small set of interlocking obligations. Each one exists to close a specific gap that the old grey zone left open.

The core provisions and what each one requires
ProvisionWhat it requires
Order-frequency thresholdAPI order flow above a specified rate, reported as roughly ten orders per second per client, is classified as algorithmic and enters the regime.
Unique algo IDEach algorithmic order carries an exchange-issued identifier, so every automated order is traceable to its strategy and provider.
Broker as principalThe broker must empanel and onboard algo and API providers as agents, register their algos, and answer for their order flow.
API access controlsTwo-factor authentication, a static or whitelisted IP, and unique client-specific and vendor-specific API keys.
White-box versus black-boxDisclosed-logic algos are lighter to run; opaque black-box offerings face heavier oversight, with providers reported to need research-analyst registration.
Exchange monitoringStock exchanges monitor compliance, register products, and carry the rule amendments that operationalise the framework.

4. The accountability chain: the broker as principal

The single most consequential design choice in the circular is where it puts responsibility. In the old model, a third-party algo or signal provider could sit close to the market with the retail user as its customer and no regulated party clearly answerable for what the software did. The framework rewires that relationship into a principal-and-agent chain with the broker at the centre.

Concretely: an algo or API provider cannot simply connect to the exchange and sell to retail users on its own. It must be empanelled by a broker, which onboards it as an agent, registers its algos, tags their orders, monitors the resulting flow and enforces controls. The broker is the principal the exchange holds to account. The retail user still uses the provider's strategy, but sits downstream of a regulated intermediary that now carries the compliance weight. That is a deliberate shift of the burden from the retail buyer, who could not police a black box, to a party the regulator already supervises.

The principal and agent accountability chain The retail user and any third-party algo or API provider connect through the broker. The broker is the principal that empanels the provider as an agent and connects to the exchange. The exchange provides oversight and monitoring, and SEBI sits above as the regulator that set the framework. Who answers for the automated order Retail user runs a strategy Algo / API provider the agent empanelled Broker the principal registers, tags, monitors Exchange oversight and monitoring SEBI · sets the framework The burden moves from the retail buyer, who cannot police a black box, to the broker the regulator already supervises.
Responsibility flows to a regulated intermediary. The provider becomes the broker's agent; the broker becomes the exchange's accountable principal. No unregistered party is left sitting between the retail user and the market with nobody answerable for the order it fires.

5. Hardening the API door

Reassigning accountability only works if the door into the broker's systems is itself controlled, so the framework tightens API access alongside the registration regime. Three controls sit at the centre of it, and they are the same controls a careful engineer would want anyway.

  • Two-factor authentication on the API session, so a leaked key alone is not enough to place orders.
  • Static or whitelisted IP. API access is bound to known network addresses, so orders cannot be sent from an arbitrary machine that has merely obtained a credential.
  • Unique API keys. Keys are issued per client and per vendor, so activity can be attributed to a specific user or provider rather than pooled behind a shared credential.

Read together with the unique algo ID, these controls make the API a metered, attributable channel rather than an open pipe. A given order can be tied to a key, the key to a client or vendor, and the algo to a registered strategy. That is the plumbing that lets the exchange trace an automated order back through the chain when it needs to.

6. White-box versus black-box

The framework treats two kinds of algo very differently, and the distinction turns on a single question: can the person trading it see the logic? A white-box algo has disclosed rules the user can inspect, understand and often configure, as with a strategy a trader builds or sets up themselves. A black-box algo keeps its logic proprietary: the user receives signals or executions but cannot verify what the strategy is actually doing.

That asymmetry is why the black box carries the heavier regime. When a provider sells outputs the buyer cannot inspect, the buyer is trusting a claim, not checking a mechanism, and the scope for opaque or overstated strategies is exactly what the circular set out to curb. Reporting on the framework indicates that black-box providers must register as a SEBI research analyst and maintain research documentation for their strategies, placing them inside a supervised, disclosure-bound perimeter. White-box activity, by contrast, largely passes the compliance weight through to the ordinary API-user path, because the logic is already visible.

White-box and black-box algos under the framework
DimensionWhite-boxBlack-box
Logic visibilityDisclosed and inspectableProprietary and hidden
What the user getsRules they can see and often configureSignals or executions only
Provider registrationLighter path; runs through the brokerReported to require research-analyst registration
Oversight weightPasses largely to the API-user pathHeavier disclosure and documentation
RationaleThe buyer can verify the mechanismThe buyer is trusting an unverifiable claim

7. The implementation timeline

A framework this structural could not switch on overnight, and it did not. The circular set a direction in February 2025, and the operative dates were then extended twice as brokers and exchanges built the systems to support registration, tagging and empanelment. The chronology below reflects the position under the 30 September 2025 extension; because timelines in this area have already moved, treat the dates as current-as-reported and confirm them against the latest SEBI and exchange notices before you rely on any one of them.

The retail algo framework implementation chronology February 2025 the circular is issued. April 2025 the standards timeline is extended. 30 September 2025 the implementation timeline is extended and ready brokers go live from 1 October 2025. 5 January 2026 brokers failing milestones are barred from onboarding new API algo clients. 1 April 2026 the full framework applies to all brokers. From circular to go-live 4 Feb 2025 Circular issued Apr 2025 Standards timeline extended 1 Oct 2025 Ready brokers may go live 5 Jan 2026 Laggards barred from new API algo clients 1 Apr 2026 Full framework, all brokers Dates reflect the 30 September 2025 extension. Timelines in this area have shifted before, so verify the current position.
A glide path, not a switch. Between the ready-broker go-live and the all-broker deadline, brokers were given interim milestones to register products and complete mock sessions, with the 5 January 2026 cut-off barring laggards from taking on new API-algo clients. The endpoint is 1 April 2026 for everyone.
The phased timeline under the 30 September 2025 extension (verify current)
DateMilestone
4 February 2025Circular Safer participation of retail investors in Algorithmic trading issued, setting the framework.
April 2025Timeline for formulating the detailed implementation standards extended.
30 September 2025Implementation timeline extended; brokers ready with the required systems may go live from 1 October 2025.
Oct to Nov 2025Glide-path milestones: brokers register retail algo products and strategies with the exchange.
5 January 2026Brokers that miss the milestones are barred from onboarding new API-based algo clients.
1 April 2026The full framework applies to all brokers, with exchanges monitoring compliance.

8. What changes for a retail algo trader

Bring it down to the person actually affected: someone running a Python strategy through a broker's API on their own capital. Three practical shifts follow from the framework, and they scale with how you trade.

If your order rate stays modest, below the threshold, you remain on the ordinary API-user path, but your access is hardened. Expect to authenticate with a second factor, to bind your access to a static or whitelisted IP, and to hold unique API keys rather than a shared credential. This is a tightening of hygiene, not a barrier.

If you cross the order-frequency threshold, your activity is algorithmic by definition, and the algo has to be registered and tagged with a unique algo ID through your broker. Your orders stop being anonymous to the exchange; they carry an identifier back to the strategy.

If you use a third-party algo or provider, that provider must be one your broker has empanelled, and an opaque black-box service brings the heavier registration burden described above. The casual, unregistered API algo bought from an unaccountable seller is precisely the thing the framework is designed to close off. For a working retail trader, the honest summary is that systematic trading on your own account remains open, but it now runs inside a documented, identifiable channel rather than a grey one. If you are mapping the end-to-end path, our companion piece on how to start algo trading in India walks the pipeline from idea to a live strategy, and the broker-API Python tutorial shows what connecting to a broker API looks like in code.

Why this matters for what you read elsewhere. Many summaries fix on a single number or a single date and treat it as settled. Two things move: the exact order-frequency threshold and how it is measured are operationalised through the exchanges, and the go-live dates have already been extended more than once. Treat any source that states a hard figure without a caveat, or omits the September 2025 extension, as potentially out of date on the points that decide how the framework applies to you.

9. Why it matters, framed neutrally

It is easy to read a registration regime as either a crackdown or a rubber stamp. It is more accurately neither. The framework is SEBI's attempt to keep the retail-algo boom accountable without banning it: to let ordinary traders automate, while making every automated order traceable and putting a regulated intermediary on the hook for the providers behind it. Whether that trade of freedom for supervision is worth it is a judgement, and this guide takes no side on it. What is not in dispute is the direction of travel, which is toward identification and accountability.

The backdrop is worth stating plainly, because it explains the regulator's posture. SEBI's own studies have repeatedly found that most individual traders in the equity derivatives segment lose money. A study released in July 2025 reported that over 91 percent of individual traders in equity derivatives made losses in FY25, with aggregate net losses of about ₹1,05,603 crore after costs. Read against that base rate, a framework that makes automated order flow supervised and traceable is a regulator tightening the plumbing around an activity where the odds already run hard against the retail participant. It is not a promise that an algo will do better; nothing here should be read that way.

That is the part the framework cannot touch. Registration makes an automated order identifiable; it does not make the strategy behind it sound. The judgement about whether an edge is real, where a position is genuinely wrong, and how large it should be, sits entirely upstream of any algo ID, and that upstream judgement is exactly what the method we teach is built around. A tagged order executing a weak plan is still a weak plan. If you want the wider landscape rather than this one circular, our overview of algorithmic trading in India sets the context, and the walkthrough of backtesting Indian equities with pandas covers the evidence step that has to come before anything goes live.

Common questions

Frequently asked questions

It brings retail algorithmic and API-based trading under a formal registration and accountability regime rather than banning it. Orders placed through a broker's API above a specified order-frequency threshold are treated as algorithmic, must be registered and tagged with a unique algo ID at the exchange, and the broker is made the principal responsible for the algo and API providers it empanels. It also hardens API access and separates disclosed white-box logic from opaque black-box strategies, which carry heavier oversight.

The framework draws the line at an order-frequency threshold, widely reported as around ten orders per second, above which orders routed through a broker's API are treated as algorithmic and must be registered and tagged. Below that rate a person is generally treated as an ordinary API user. The exact figure and how it is measured are set through the exchanges and the industry standards process, so treat the number as operative but verify the current value with your broker and exchange before relying on it.

A unique algo ID is an exchange-issued identifier attached to every order an algorithm places through a broker's API. It lets the exchange trace each automated order back to the specific strategy and provider that generated it, creating an audit trail across the whole chain. The broker registers the algo with the exchange and ensures orders carry the tag, so identification is built into the order itself rather than reconstructed after the fact.

It means accountability sits with the broker, not only with the software vendor. Under the framework the broker is the principal that must empanel or onboard algorithmic and API providers as its agents, register their algos, monitor order flow and enforce controls. Providers cannot connect straight to the exchange on their own; they operate through a broker that answers for them. That reassigns responsibility for retail algo activity to a regulated intermediary the exchange can hold to account.

A white-box algo has disclosed logic: the user can see and understand the rules that generate its orders, as with a strategy a person configures themselves. A black-box algo keeps its logic proprietary, so the user receives only signals or executions and cannot independently verify the strategy. The framework treats black-box offerings more strictly: reporting indicates their providers must register as a SEBI research analyst and maintain research documentation, because the buyer cannot inspect what they are trading.

The circular was issued on 4 February 2025 and its timeline was extended twice. Under the 30 September 2025 extension, brokers ready with the required systems could go live from 1 October 2025, brokers were given glide-path milestones through late 2025, brokers failing those milestones were to be barred from onboarding new API-based algo clients from 5 January 2026, and the full framework was set to apply to all brokers from 1 April 2026. Confirm the current dates, since timelines in this area have shifted before.

If your order rate stays below the threshold you are generally treated as an ordinary API user, but your access is hardened: two-factor authentication, a static or whitelisted IP, and unique API keys. If you cross the threshold, or if you use a third-party algo provider, the algo has to be registered and tagged with a unique algo ID through your broker, and any provider you use must be one the broker has empanelled. Practically, the era of anonymous, unregistered API algos is closing.

No. Its stated aim is safer participation of retail investors in algorithmic trading, not prohibition. It keeps the door open to retail algos while adding identification, empanelment and accountability so that automated order flow is traceable and someone regulated answers for it. The intent is to let the retail-algo boom continue under supervision rather than to shut it down, which is why the framing is a registration regime rather than a ban.

Broker APIs and no-code platforms made algorithmic execution accessible to ordinary retail traders, and a market of third-party algo and signal providers grew around them with little formal oversight, including opaque strategies sold on unverifiable claims. The framework responds by making automated orders identifiable, placing accountability on the broker, and separating disclosed logic from black boxes. It sits against a backdrop where SEBI's own studies show most individual derivatives traders lose money, so the regulator is tightening the plumbing around a high-risk activity.

Where the facts come from

Sources

  • SEBI circular, 4 February 2025. Safer participation of retail investors in Algorithmic trading, reference SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/0000013, which sets the framework: the classification of API order flow as algorithmic, the unique algo ID, the broker-as-principal empanelment model, API access controls, and the white-box versus black-box treatment. sebi.gov.in
  • SEBI extension circular, 30 September 2025. Extends the implementation timeline: ready brokers may go live from 1 October 2025, glide-path milestones run through late 2025, laggard brokers are barred from onboarding new API-algo clients from 5 January 2026, and the full framework applies to all brokers from 1 April 2026. sebi.gov.in
  • SEBI study on individual derivatives traders, July 2025. Reported that over 91 percent of individual traders in equity derivatives made losses in FY25, with aggregate net losses of about ₹1,05,603 crore after transaction costs, the base rate against which the framework is set. business-standard.com
  • Exchange framework material. Stock-exchange FAQs and notices operationalise the order-frequency threshold, the unique algo ID, empanelment and the mock-session milestones; the precise threshold value and its measurement are set through the exchanges and the industry standards process, so verify the current figure with your broker and exchange.
Educational note. This guide explains SEBI's retail algo-trading framework and its mechanics. It is not a recommendation to trade, to automate, or to buy or sell any security, and it is not investment or legal advice. Bharath Shiksha is an educational publisher, not a SEBI-registered investment adviser or research analyst. Regulatory dates and thresholds in this area have changed before; confirm the current position against primary SEBI and exchange notices before relying on any specific figure or date.

Related guides

Registration makes the order traceable. The plan behind it is the harder half.